Technical results

require(readr)

## Loading required package: readr

require(dplyr)

## Loading required package: dplyr

## Warning: package 'dplyr' was built under R version 3.4.2

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

require(stringr)

## Loading required package: stringr

require(zipcode)

## Loading required package: zipcode

## Warning in library(package, lib.loc = lib.loc, character.only = TRUE,
## logical.return = TRUE, : there is no package called 'zipcode'

require(tidyverse)

## Loading required package: tidyverse

## Loading tidyverse: ggplot2
## Loading tidyverse: tibble
## Loading tidyverse: tidyr
## Loading tidyverse: purrr

## Conflicts with tidy packages ----------------------------------------------

## filter(): dplyr, stats
## lag():    dplyr, stats

require(ggmap)

## Loading required package: ggmap

require(lubridate)

## Loading required package: lubridate

## Warning: package 'lubridate' was built under R version 3.4.2

## 
## Attaching package: 'lubridate'

## The following object is masked from 'package:base':
## 
##     date

require(viridis)

## Loading required package: viridis

## Loading required package: viridisLite

require(ggthemes)

## Loading required package: ggthemes

require(plotly)

## Loading required package: plotly

## 
## Attaching package: 'plotly'

## The following object is masked from 'package:ggmap':
## 
##     wind

## The following object is masked from 'package:ggplot2':
## 
##     last_plot

## The following object is masked from 'package:stats':
## 
##     filter

## The following object is masked from 'package:graphics':
## 
##     layout

require(reshape2)

## Loading required package: reshape2

## 
## Attaching package: 'reshape2'

## The following object is masked from 'package:tidyr':
## 
##     smiths

nypd.vehicle <- read_csv("NYPD_Motor_Vehicle_Collisions.csv", col_names = T)

## Parsed with column specification:
## cols(
##   .default = col_character(),
##   TIME = col_time(format = ""),
##   `ZIP CODE` = col_integer(),
##   LATITUDE = col_double(),
##   LONGITUDE = col_double(),
##   `NUMBER OF PERSONS INJURED` = col_integer(),
##   `NUMBER OF PERSONS KILLED` = col_integer(),
##   `NUMBER OF PEDESTRIANS INJURED` = col_integer(),
##   `NUMBER OF PEDESTRIANS KILLED` = col_integer(),
##   `NUMBER OF CYCLIST INJURED` = col_integer(),
##   `NUMBER OF CYCLIST KILLED` = col_integer(),
##   `NUMBER OF MOTORIST INJURED` = col_integer(),
##   `NUMBER OF MOTORIST KILLED` = col_integer(),
##   `UNIQUE KEY` = col_integer()
## )

## See spec(...) for full column specifications.

summary(nypd.vehicle)

##      DATE               TIME            BOROUGH             ZIP CODE     
##  Length:1152443     Length:1152443    Length:1152443     Min.   :10000   
##  Class :character   Class1:hms        Class :character   1st Qu.:10128   
##  Mode  :character   Class2:difftime   Mode  :character   Median :11205   
##                     Mode  :numeric                       Mean   :10812   
##                                                          3rd Qu.:11236   
##                                                          Max.   :11697   
##                                                          NA's   :323606  
##     LATITUDE        LONGITUDE         LOCATION         ON STREET NAME    
##  Min.   : 0.00    Min.   :-201.36   Length:1152443     Length:1152443    
##  1st Qu.:40.67    1st Qu.: -73.98   Class :character   Class :character  
##  Median :40.72    Median : -73.93   Mode  :character   Mode  :character  
##  Mean   :40.72    Mean   : -73.92                                        
##  3rd Qu.:40.77    3rd Qu.: -73.87                                        
##  Max.   :41.13    Max.   :   0.00                                        
##  NA's   :211564   NA's   :211564                                         
##  CROSS STREET NAME  OFF STREET NAME    NUMBER OF PERSONS INJURED
##  Length:1152443     Length:1152443     Min.   : 0.0000          
##  Class :character   Class :character   1st Qu.: 0.0000          
##  Mode  :character   Mode  :character   Median : 0.0000          
##                                        Mean   : 0.2567          
##                                        3rd Qu.: 0.0000          
##                                        Max.   :43.0000          
##                                                                 
##  NUMBER OF PERSONS KILLED NUMBER OF PEDESTRIANS INJURED
##  Min.   :0.000000         Min.   : 0.00000             
##  1st Qu.:0.000000         1st Qu.: 0.00000             
##  Median :0.000000         Median : 0.00000             
##  Mean   :0.001201         Mean   : 0.05209             
##  3rd Qu.:0.000000         3rd Qu.: 0.00000             
##  Max.   :8.000000         Max.   :28.00000             
##                                                        
##  NUMBER OF PEDESTRIANS KILLED NUMBER OF CYCLIST INJURED
##  Min.   :0.000000             Min.   :0.00000          
##  1st Qu.:0.000000             1st Qu.:0.00000          
##  Median :0.000000             Median :0.00000          
##  Mean   :0.000664             Mean   :0.02078          
##  3rd Qu.:0.000000             3rd Qu.:0.00000          
##  Max.   :8.000000             Max.   :4.00000          
##                                                        
##  NUMBER OF CYCLIST KILLED NUMBER OF MOTORIST INJURED
##  Min.   :0.00e+00         Min.   : 0.0000           
##  1st Qu.:0.00e+00         1st Qu.: 0.0000           
##  Median :0.00e+00         Median : 0.0000           
##  Mean   :7.98e-05         Mean   : 0.1852           
##  3rd Qu.:0.00e+00         3rd Qu.: 0.0000           
##  Max.   :2.00e+00         Max.   :43.0000           
##                                                     
##  NUMBER OF MOTORIST KILLED CONTRIBUTING FACTOR VEHICLE 1
##  Min.   :0.000000          Length:1152443               
##  1st Qu.:0.000000          Class :character             
##  Median :0.000000          Mode  :character             
##  Mean   :0.000458                                       
##  3rd Qu.:0.000000                                       
##  Max.   :5.000000                                       
##                                                         
##  CONTRIBUTING FACTOR VEHICLE 2 CONTRIBUTING FACTOR VEHICLE 3
##  Length:1152443                Length:1152443               
##  Class :character              Class :character             
##  Mode  :character              Mode  :character             
##                                                             
##                                                             
##                                                             
##                                                             
##  CONTRIBUTING FACTOR VEHICLE 4 CONTRIBUTING FACTOR VEHICLE 5
##  Length:1152443                Length:1152443               
##  Class :character              Class :character             
##  Mode  :character              Mode  :character             
##                                                             
##                                                             
##                                                             
##                                                             
##    UNIQUE KEY      VEHICLE TYPE CODE 1 VEHICLE TYPE CODE 2
##  Min.   :     22   Length:1152443      Length:1152443     
##  1st Qu.: 290312   Class :character    Class :character   
##  Median :3212342   Mode  :character    Mode  :character   
##  Mean   :2286022                                          
##  3rd Qu.:3500480                                          
##  Max.   :3789579                                          
##                                                           
##  VEHICLE TYPE CODE 3 VEHICLE TYPE CODE 4 VEHICLE TYPE CODE 5
##  Length:1152443      Length:1152443      Length:1152443     
##  Class :character    Class :character    Class :character   
##  Mode  :character    Mode  :character    Mode  :character   
##                                                             
##                                                             
##                                                             
## 

1) The most dangerous/safest intersections in NYC

First, we mutate two new variables Num_Injured and Num_Killed which are the sum of all injured or killed people (people in the vehicles, pedestrians, cyclists, motorists) and then group them by cross street. We then summarise by number of accidents, number of injuries and number of deaths for each of the cross streets.

colnames(nypd.vehicle) <- str_replace_all(colnames(nypd.vehicle), " ", "_")
q1 <- nypd.vehicle %>% 
  mutate(num = 1,
         Num_Injured = NUMBER_OF_PERSONS_INJURED + NUMBER_OF_PEDESTRIANS_INJURED + NUMBER_OF_CYCLIST_INJURED + NUMBER_OF_MOTORIST_INJURED,
         Num_Killed = NUMBER_OF_PERSONS_KILLED + NUMBER_OF_PEDESTRIANS_KILLED + NUMBER_OF_CYCLIST_KILLED + NUMBER_OF_MOTORIST_KILLED
         ) %>%
  filter(!is.na(CROSS_STREET_NAME)) %>%
  group_by(CROSS_STREET_NAME) %>%
  summarise(Number_of_Accidents = n(), 
            Num_Injured = sum(Num_Injured),
            Num_Killed = sum(Num_Killed))

We use three different criteria to determine the most dangerous/safest intersections based on Number_of_Accidents, Num_Injured, and Num_Killed. 3rd ave was the most dangerous in terms of accidents and Broadway in terms of number injured and killed. We created tables for the safest and most dangerous intersections, using max of accidents and min of accidents, respectively. Non-busy places such as precincts have the lowest number of non-zero accidents.

knitr::kable(q1[q1$Number_of_Accidents == max(q1$Number_of_Accidents), ], 
             caption = "Most dangerous NYC intersections")

Most dangerous NYC intersections

CROSS_STREET_NAME	Number_of_Accidents	Num_Injured	Num_Killed
3 AVENUE	12116	4547	10

knitr::kable(q1[q1$Num_Injured == max(q1$Num_Injured), ], 
             caption = "Most dangerous NYC intersections")

Most dangerous NYC intersections

CROSS_STREET_NAME	Number_of_Accidents	Num_Injured	Num_Killed
BROADWAY	11871	4714	28

knitr::kable(q1[q1$Num_Killed == max(q1$Num_Killed), ], 
             caption = "Most dangerous NYC intersections")

Most dangerous NYC intersections

CROSS_STREET_NAME	Number_of_Accidents	Num_Injured	Num_Killed
BROADWAY	11871	4714	28

safe <- q1[q1$Number_of_Accidents == min(q1$Number_of_Accidents) & q1$Num_Killed == 0 &
              q1$Num_Injured == 0,]
knitr::kable(head(safe), caption = "Safest NYC intersections")

Safest NYC intersections

CROSS_STREET_NAME	Number_of_Accidents	Num_Injured	Num_Killed
0	1	0	0
043 PCT	1	0	0
1 ave	1	0	0
1 Avenue	1	0	0
1 AVENUE + 2 AVENUE	1	0	0
1 AVENUE TO PLEASANT AVENUE	1	0	0

We then made 2 bar graphs out of this to plot the top 5 most dangerous and safest NYC cross streets, each broken down into 3 bars by Number_of_Accidents, Num_Injured, and Num_Killed. We did this through the use of gather and dodge positioning.

ggplot(data = q1 %>% arrange(-Number_of_Accidents) %>% top_n(5,Number_of_Accidents) %>% 
         gather("Number_of_Accidents", "Num_Injured", "Num_Killed",key = Type, value = Amount),
       aes(x = CROSS_STREET_NAME, y =  Amount, fill = Type)) +
   geom_col(position = "dodge") +
  ggtitle("Most dangerous NYC intersections") + xlab("Cross street") + ylab("number") +
  theme(axis.text.x  = element_text(angle = 0),
        text = element_text(size = 10, hjust=0.5 , vjust = 1, face = "bold"))

From the plot, we can direcly find that the intersection which has the greatest number of injured people is Broadway and the one which has the greatest number of accidents is 3 Avenue, but the number of killed people is hard to see because of its scale relative to the accident and injured quantity. If you look closely though, Broadway definitely has the thickest and most visible green mark for number killed.

ggplot(data = q1 %>% arrange(Number_of_Accidents)  %>% head(5) %>% 
         gather("Number_of_Accidents", "Num_Injured", "Num_Killed",key = Type, value = Amount),
       aes(x = CROSS_STREET_NAME, y =  Amount, fill = Type)) +
   geom_col(position = "dodge") +
  ggtitle("Safest NYC intersections") + xlab("Cross street") + ylab("number") +
  theme(axis.text.x  = element_text(angle = 0),
        text = element_text(size = 10, hjust=0.5, vjust = 1, face = "bold"))

From the plot, we can direcly find that the intersections which have the least number of injured people is 0, 043 PCT and 1 Avenue, and all of the top 5 safest intersections have "1" for number of accidents and "0" for the number of killed people.

2) Distribution of killed people based on the map of NYC

From the plot of the most dangerous intersections of NYC, we noticed that the number of killed people is to small to see from the plot. We use ggmap to get the google map of NYC and figure out the distribution of people killed based on the map of NYC and utilizing the latitude and longitude coordinates. We created a data frame that gave the lat. and long. for each accident in which someone was killed, by omitting ones in which no one was killed with filter.

colnames(nypd.vehicle) <- str_replace_all(colnames(nypd.vehicle), " ", "_")
nypd <- nypd.vehicle %>%
  select(LATITUDE,LONGITUDE,NUMBER_OF_PERSONS_KILLED) %>% 
  gather(type,value,3:3) %>%
  na.omit() %>% 
  group_by(LATITUDE,LONGITUDE,type) %>% 
  summarise(total=sum(value,na.rm=T)) %>% 
  filter(total!=0)
nypd

## # A tibble: 996 x 4
## # Groups:   LATITUDE, LONGITUDE [996]
##    LATITUDE LONGITUDE                     type total
##       <dbl>     <dbl>                    <chr> <int>
##  1 40.50619 -74.23489 NUMBER_OF_PERSONS_KILLED     1
##  2 40.51387 -74.23668 NUMBER_OF_PERSONS_KILLED     1
##  3 40.51662 -74.20279 NUMBER_OF_PERSONS_KILLED     1
##  4 40.52400 -74.18610 NUMBER_OF_PERSONS_KILLED     1
##  5 40.52664 -74.16813 NUMBER_OF_PERSONS_KILLED     1
##  6 40.52937 -74.16135 NUMBER_OF_PERSONS_KILLED     2
##  7 40.53078 -74.22443 NUMBER_OF_PERSONS_KILLED     2
##  8 40.53427 -74.15397 NUMBER_OF_PERSONS_KILLED     2
##  9 40.53866 -74.18800 NUMBER_OF_PERSONS_KILLED     1
## 10 40.54101 -74.14710 NUMBER_OF_PERSONS_KILLED     1
## # ... with 986 more rows

nyc <- get_map("new york",zoom=11)

## Map from URL : http://maps.googleapis.com/maps/api/staticmap?center=new+york&zoom=11&size=640x640&scale=2&maptype=terrain&language=en-EN&sensor=false

## Information from URL : http://maps.googleapis.com/maps/api/geocode/json?address=new%20york&sensor=false

map_killed <- ggmap(nyc)+geom_point(data=subset(nypd,type=="NUMBER_OF_PERSONS_KILLED"), 
             aes(x=LONGITUDE, y=LATITUDE, colour=total),size= 2,alpha=0.2) +
  ggtitle("Distribution of deaths")+xlab("Longitude")+ylab("Latitude") + 
  scale_color_gradient(low = "red",  high = "black")
map_killed

## Warning: Removed 125 rows containing missing values (geom_point).

From the map, persons who got killed in the collisions are relatively few compared to persons who got injured. Most of the collisions involving death of people took place in Manhattan, Brooklyn and Queens and then Bronx, while Staten island has the least number of death of people.

3) The probability that people were injured or killed from the collision

We combined the number of people injured and the number of people killed to figure out the proportion of crashes that resulted in an injury or death. We got the proportion by taking the mean of the conditional statement of whether someone got injured or killed.

q4 <- nypd.vehicle %>%
  mutate(Injured_or_Killed = NUMBER_OF_PERSONS_INJURED + NUMBER_OF_PEDESTRIANS_INJURED + NUMBER_OF_CYCLIST_INJURED + NUMBER_OF_MOTORIST_INJURED +NUMBER_OF_PERSONS_KILLED + NUMBER_OF_PEDESTRIANS_KILLED + NUMBER_OF_CYCLIST_KILLED + NUMBER_OF_MOTORIST_KILLED
  ) %>%
  mutate(Injured_or_Killed_Yes = ifelse(Injured_or_Killed > 0, 1, 0))


prop <- mean(q4$Injured_or_Killed_Yes)
prop

## [1] 0.1886957

It was around 19% of people that get injured or killed per crash.

4) Distribution of collisions by boroughs

In order to look at how the average number of accidents varied over time within each of the 5 boroughs throughout the years of the dataset, we filtered the data by the borough variable and then used group_by to organize by the date of the year and borough and summarised an average accidents variable that was the mean of the number of accidents per month in that borough at each of those moments in time. We did this with a geom_line to track the trends and ebbs and flows for each and grouped by the date it occured and the borough it occurred in and summarise the average collisions per month for those groups.

nypd.vehicle$DATE <- mdy(nypd.vehicle$DATE)
borough.accident <- nypd.vehicle %>% filter(BOROUGH!="") %>%  group_by(DATE,BOROUGH) %>% summarise(Average_CollisionsPerMonth = mean(n())) %>% na.omit()

ggplot(data=borough.accident,aes(x=DATE, y=Average_CollisionsPerMonth, colour=BOROUGH, group=BOROUGH)) + geom_line()+geom_point(size=0.5,shape=1) + ggtitle("Average Accidents per month by Borough")+geom_text(aes(label=ifelse(Average_CollisionsPerMonth>180,Average_CollisionsPerMonth,"")), size=2,hjust=1.5)

The result seems quite obvious.
The overall ranking of the boroughs from the most accidents happened to the least is:
1. Brooklyn
2. Queens
3. Manhattan
4. Bronx
5. Staten Island

5) Type of vehicles

In order to find the type of vehicles (pickup truck, passenger, taxi, etc.) most frequently caused a collision, we add the two vehicle types together and sort the 10 most common types of vehicles based on the number of collisions. We did this by creating a union of vehicle_1 and vehicle_2 from q5.1 and q5.2, as there are usually 2 vehicles involved in a collision.

q5.1 <- nypd.vehicle %>%
  rename(VEHICLE_TYPE = VEHICLE_TYPE_CODE_1) %>%
  group_by(VEHICLE_TYPE) %>%
  summarise(Number_of_Accidents = n()) %>%
  mutate(Type = "VEHICLE_TYPE_CODE_1")
  
q5.2 <- nypd.vehicle %>%
  rename(VEHICLE_TYPE = VEHICLE_TYPE_CODE_2) %>%
  group_by(VEHICLE_TYPE) %>%
  summarise(Number_of_Accidents = n()) %>%
  mutate(Type = "VEHICLE_TYPE_CODE_2")

q2 <- union_all(q5.1, q5.2) %>%
  filter(!VEHICLE_TYPE %in% c("UNKNOWN", "OTHER", NA)) %>%
  group_by(Type) %>%
  arrange(-Number_of_Accidents) %>%
  top_n(10,Number_of_Accidents)
knitr::kable(q2, caption = "Number of Collisions by Vehicle Type")

Number of Collisions by Vehicle Type

VEHICLE_TYPE	Number_of_Accidents	Type
PASSENGER VEHICLE	662184	VEHICLE_TYPE_CODE_1
PASSENGER VEHICLE	496070	VEHICLE_TYPE_CODE_2
SPORT UTILITY / STATION WAGON	274688	VEHICLE_TYPE_CODE_1
SPORT UTILITY / STATION WAGON	205959	VEHICLE_TYPE_CODE_2
TAXI	44544	VEHICLE_TYPE_CODE_1
TAXI	36640	VEHICLE_TYPE_CODE_2
VAN	26846	VEHICLE_TYPE_CODE_1
BICYCLE	25180	VEHICLE_TYPE_CODE_2
VAN	23905	VEHICLE_TYPE_CODE_2
PICK-UP TRUCK	20446	VEHICLE_TYPE_CODE_1
PICK-UP TRUCK	17816	VEHICLE_TYPE_CODE_2
SMALL COM VEH(4 TIRES)	14874	VEHICLE_TYPE_CODE_1
SMALL COM VEH(4 TIRES)	14627	VEHICLE_TYPE_CODE_2
LARGE COM VEH(6 OR MORE TIRES)	14117	VEHICLE_TYPE_CODE_1
BUS	14044	VEHICLE_TYPE_CODE_1
LARGE COM VEH(6 OR MORE TIRES)	13367	VEHICLE_TYPE_CODE_2
BUS	11460	VEHICLE_TYPE_CODE_2
LIVERY VEHICLE	9615	VEHICLE_TYPE_CODE_1
LIVERY VEHICLE	7704	VEHICLE_TYPE_CODE_2
MOTORCYCLE	6607	VEHICLE_TYPE_CODE_1

We then created a bar graph that illustrated the 10 most frequent car types that got into a collision, broken down by the 2 vehicle codes and how many collisions there were involving those vehicle types.

ggplot(data = q2,
       aes(x = VEHICLE_TYPE, y =  Number_of_Accidents, fill = Type)) +
   geom_col(position = "stack") +
  ggtitle("Number of Collisions by Vehicle Type") + ylab("Number of collisions") + xlab("Vehicle_Type") +
  theme(axis.text.x  = element_text(angle = 0),
        text = element_text(size = 10, hjust=0.5, vjust = 1, face = "bold")) + coord_flip()

From the plot, passanger vehicles are involved in much more accidents than any other vehicles. Sport utility/station wagon come in second, involved in less accidents than the passanger vehicles but still a lot relative to even the third place vehicle. The rest seems similar after taxis.

6) The most popular contributing factors

Then we try to find the most common contributing factors leading to a collision. We group together the factors similar to what we did with the vehicle types for vehicle 1 and vehicle 2 in the accident and show the top ten factors that led to a crash.

q3.1 <- nypd.vehicle %>%
  rename(CONTRIBUTING_FACTOR = CONTRIBUTING_FACTOR_VEHICLE_1) %>%
  group_by(CONTRIBUTING_FACTOR) %>%
  summarise(Number_of_Accidents = n()) %>%
  mutate(Type = "CONTRIBUTING_FACTOR_VEHICLE_1")
  
q3.2 <- nypd.vehicle %>%
  rename(CONTRIBUTING_FACTOR = CONTRIBUTING_FACTOR_VEHICLE_2) %>%
  group_by(CONTRIBUTING_FACTOR) %>%
  summarise(Number_of_Accidents = n()) %>%
  mutate(Type = "CONTRIBUTING_FACTOR_VEHICLE_2")

q3 <- union_all(q3.1, q3.2) %>%
  filter(!CONTRIBUTING_FACTOR %in% c("UNKNOWN", "Unspecified", NA)) %>%
  group_by(Type) %>%
  arrange(-Number_of_Accidents) %>%
  top_n(10,Number_of_Accidents)
knitr::kable(q3, caption = "Number of Collisions for each Contributing Factor")

Number of Collisions for each Contributing Factor

CONTRIBUTING_FACTOR	Number_of_Accidents	Type
Driver Inattention/Distraction	166389	CONTRIBUTING_FACTOR_VEHICLE_1
Failure to Yield Right-of-Way	54010	CONTRIBUTING_FACTOR_VEHICLE_1
Fatigued/Drowsy	48457	CONTRIBUTING_FACTOR_VEHICLE_1
Driver Inattention/Distraction	46476	CONTRIBUTING_FACTOR_VEHICLE_2
Backing Unsafely	35893	CONTRIBUTING_FACTOR_VEHICLE_1
Other Vehicular	34846	CONTRIBUTING_FACTOR_VEHICLE_1
Turning Improperly	28209	CONTRIBUTING_FACTOR_VEHICLE_1
Following Too Closely	27306	CONTRIBUTING_FACTOR_VEHICLE_1
Lost Consciousness	20232	CONTRIBUTING_FACTOR_VEHICLE_1
Other Vehicular	20047	CONTRIBUTING_FACTOR_VEHICLE_2
Prescription Medication	15900	CONTRIBUTING_FACTOR_VEHICLE_1
Traffic Control Disregarded	15417	CONTRIBUTING_FACTOR_VEHICLE_1
Fatigued/Drowsy	13027	CONTRIBUTING_FACTOR_VEHICLE_2
Failure to Yield Right-of-Way	10393	CONTRIBUTING_FACTOR_VEHICLE_2
Turning Improperly	6620	CONTRIBUTING_FACTOR_VEHICLE_2
Lost Consciousness	6344	CONTRIBUTING_FACTOR_VEHICLE_2
Backing Unsafely	5674	CONTRIBUTING_FACTOR_VEHICLE_2
Driver Inexperience	4326	CONTRIBUTING_FACTOR_VEHICLE_2
Following Too Closely	4019	CONTRIBUTING_FACTOR_VEHICLE_2
Traffic Control Disregarded	3831	CONTRIBUTING_FACTOR_VEHICLE_2

We then created a bar graph that illustrated the 10 most popular contributing factors that got into a collision, broken down by the 2 vehicle codes and the amount of crashes these factors were involved in.

ggplot(data = q3,
       aes(x = CONTRIBUTING_FACTOR, y =  Number_of_Accidents, fill = Type)) +
   geom_col(position = "stack") +
  ggtitle("Number of Collisions by Contributing Factor") + xlab("Contributing Factor") +
  theme(axis.text.x  = element_text(angle = 0),
        text = element_text(size = 10, hjust=0.5, vjust = 1, face = "bold")) + coord_flip()

From the result, driver inattention/distraction is the most common contributing factors that caused collisions. This is because driver inattention can be looking off the road for a second, texting and driving, eating, etc. We would think failure to yield the right of way would be more common in a city where everyone is in a hurry.

7) Time-interval

In order to figure out certain time periods that the collision occured more frequently, we split one day into 24 hours and count the number of collisions happened in each time interval.

# 60 minute
q5.1 <- nypd.vehicle %>%
  group_by(TIME) %>%
  summarise(n = n()) %>%
  mutate(time = hms(TIME),
         time = hour(time)*3600 + minute(time)*60 + second(time),
         times = ceiling(time/3600),
         time.end = time/3600
         )

q5.interval <- q5.1 %>%
  filter(times == time.end) %>%
  union_all(q5.1 %>% top_n(1)) %>%
  select(times, TIME)

## Selecting by time.end

q5.interval <- q5.interval %>%
  rename(start_time = TIME) %>%
  inner_join(
    q5.interval %>%
      rename(end_time = TIME) %>%
  mutate(times = times - 1) 
  )

## Joining, by = "times"

q5.2 <- q5.1 %>%
  group_by(times) %>%
  summarise(n = sum(n)) %>%
  inner_join(q5.interval) %>%
  arrange(-n) %>% 
  mutate(Collisions = n) %>%
  select(start_time, end_time, Collisions) %>%
  mutate(time_interval = paste0(start_time, "->", end_time)) %>%
  arrange(start_time)

## Joining, by = "times"

knitr::kable(q5.2, caption = "Most frequent 1-hour intervals for collisions to occur")

Most frequent 1-hour intervals for collisions to occur

start_time	end_time	Collisions	time_interval
00:00:00	01:00:00	4683	00:00:00->01:00:00
01:00:00	02:00:00	28900	01:00:00->02:00:00
02:00:00	03:00:00	17467	02:00:00->03:00:00
03:00:00	04:00:00	13518	03:00:00->04:00:00
04:00:00	05:00:00	11826	04:00:00->05:00:00
05:00:00	06:00:00	14170	05:00:00->06:00:00
06:00:00	07:00:00	15668	06:00:00->07:00:00
07:00:00	08:00:00	25501	07:00:00->08:00:00
08:00:00	09:00:00	37082	08:00:00->09:00:00
09:00:00	10:00:00	66095	09:00:00->10:00:00
10:00:00	11:00:00	63088	10:00:00->11:00:00
11:00:00	12:00:00	59013	11:00:00->12:00:00
12:00:00	13:00:00	62297	12:00:00->13:00:00
13:00:00	14:00:00	66217	13:00:00->14:00:00
14:00:00	15:00:00	69425	14:00:00->15:00:00
15:00:00	16:00:00	80642	15:00:00->16:00:00
16:00:00	17:00:00	71858	16:00:00->17:00:00
17:00:00	18:00:00	86260	17:00:00->18:00:00
18:00:00	19:00:00	82322	18:00:00->19:00:00
19:00:00	20:00:00	70778	19:00:00->20:00:00
20:00:00	21:00:00	57371	20:00:00->21:00:00
21:00:00	22:00:00	47917	21:00:00->22:00:00
22:00:00	23:00:00	40058	22:00:00->23:00:00
23:00:00	23:59:00	36018	23:00:00->23:59:00

ggplot(data = q5.2,
       aes(x = seq(0,23,1), y =  Collisions, fill = time_interval)) +
   geom_col(position = "stack") +
  ggtitle("Collisions by Hour") + xlab("Time(hrs)") + ylab("Number of collisions")+
    theme(axis.text.x  = element_text(angle = 0),
        text = element_text(hjust=0.5, vjust = 1, size = 10, face = "bold"),
        legend.position="none")

From the plot, we found that the number of accidents rises suddenly between 8am and 9am and continues to rise through 3 pm. The maximum hour in terms of accidents was at 5pm.